**** BEGIN LOGGING AT Tue Jul 24 03:00:01 2018 Jul 24 07:42:58 New news from stackoverflow: Why in final image of yocto asterisk is not available? Jul 24 08:21:39 Hello there, I discovered a problem in the kernel-devsrc.bb recipe (rocko branch), how should I deliver my patch? Jul 24 10:43:29 New news from stackoverflow: Using files generated by other recipes in Yocto Jul 24 10:49:33 khem: https://git.dereferenced.org/pkgconf/pkgconf/issues/2 Jul 24 10:50:34 khem: also https://git.dereferenced.org/pkgconf/pkgconf/commit/41f8031533f182ba04b2c5848b75f0d7707df944 Jul 24 12:29:54 Hello everybody! I'm experiencing some issues when trying to compile PostGIS 2.2.7 (PostgreSQL 9.4.18 installed from OpenEmebedded) with Yocto Jul 24 12:31:26 The do_configure can't be completed because pg_config (postgresql) can't be accessed by the postgis recipe...that file is located only in the workdir of postgresql Jul 24 12:31:55 More info here: https://stackoverflow.com/questions/51496061/using-files-generated-by-other-recipes-in-yocto Jul 24 12:51:52 Also, I have tried pointing --with-pgconfig to that pg_config (postgresql work folder), but than fails to get the library paths... Jul 24 12:53:35 Is there a way to make that binaries available to other recipes? Jul 24 12:53:50 those* Jul 24 14:25:39 With a read-only rootfs, is there a cleaner way to make network-related /etc files be volatile (like /etc/hostname and /etc/network/interfaces)? Jul 24 14:25:59 Cleaner than extending the initscripts "volatiles" file, I meant to say. Jul 24 14:30:20 tgoodwin: you use sysvinit or systemd? Jul 24 14:31:39 oops iniscripts so i assume sysvinit :) you can switch to systemd and use volatile-bind, or you can add bind mount by hand at rootfs creation Jul 24 14:31:48 looks like sysvinit (poky) Jul 24 14:34:10 nayfe: thanks. I was looking at the initscripts recipe since it looks like I could toss a file into /etc/default/volatiles/00_core and it would get picked up by populate-volatile.sh. Jul 24 14:35:00 But I wasn't terribly confident in my understanding of that recipe since on boot I see some messages like: rm can't remove /etc/resolv.conf, etc. Jul 24 14:41:05 tgoodwin apparently you can use bind mount with populate-volatile.sh https://github.com/openembedded/openembedded-core/blob/master/meta/recipes-core/initscripts/initscripts-1.0/populate-volatile.sh#L179, can you show your modified 00_core? Jul 24 14:42:27 Right, that's what I'm working towards right now. That script seems to read any files in 00_core, so my plan was to put another file in it containing for example "l root root 0644 /etc/hostname /var/run/hostname" Jul 24 14:44:14 New news from stackoverflow: libwayland-egl.so is installed by both wayland and mali-t62x Jul 24 14:47:10 I'll let you know if this works once the build finishes. Jul 24 14:47:41 Thanks for the suggestions. Would there be any greater overall benefit to using systemd vs. sysvinit given that the latter seems to be the poky default? Jul 24 14:47:57 (beyond just the volatile files/directories of course) Jul 24 14:51:34 you can try "b root root 0644 /etc/hostname /var/run/hostname" instead (l -> b) otherwise Jul 24 14:53:19 systemd has been chosen by many distro :p you can find a lot on systemd vs sysvinit debate :) Jul 24 14:55:42 naturally ;-) Jul 24 16:27:51 are the -c clean and -c cleansstate options valid for an image? Jul 24 16:28:33 they'll clean out the workdir of the image recipe, that's it Jul 24 16:29:30 so do i have to "rm -fR /cache-sstate" to get the tmp/deploy/images/xxx stuff to rebuild?!? Jul 24 16:30:29 i create a new set of patches in my layer (for connman) that don't seem to be picked up with a basic "bitbake hw-test-image" Jul 24 16:30:34 s/create/created/ Jul 24 16:31:38 they should be, right? i shouldn't have to remove cache-sstate, should i? Jul 24 16:33:26 no, it did. i misread the file timestamps. Jul 24 16:34:39 rm -rf tmp; bitbake —no-setscene some-image Jul 24 16:35:15 but no, tmp/deploy/images/ stuff will rebuild fine if you clean the image, you're right, though remember there are other recipes that put files there too, i.e. kernel Jul 24 16:35:19 really depends on your gaols Jul 24 16:35:40 if all you want is to reconstruct the image, bitbake -C rootfs some-image Jul 24 16:39:47 yates: if you need to delete stuff in sstate-cache to make it rebuild then chances are bitbake doesn't know about your change Jul 24 17:33:31 Does anyone remember the u-boot race condition we were talking about last Thursday? I think I found a fix upstream and I thought there was a Bugzilla for it, and I cannot seem to find it now Jul 24 17:41:32 nayfe: we had the right idea for sysvinit. Create a file like 99_my_extra_volatiles over in sysconfdir/default/volatiles with your extra links/files/directories and it all works fine. The "b" option wasn't listed in the original volatiles file so I didn't try to use it. Just link (l) followed by file (f) to none. Jul 24 17:55:50 ok, thanks kergoth/rbu Jul 24 17:55:54 rburton Jul 24 21:13:46 is it possible to increase the size of the boot partition when bitbaking an image? Jul 24 21:16:45 depends on the image construction method. if your bsp is using wic, you can switch to a different wks Jul 24 21:24:29 i'm looking for a quick method of updating the kernel. I thought I could maybe copy the update to the bootfs and then replace the kernel in u-boot... Jul 24 21:25:47 or is there a more established method? Jul 24 21:27:02 method will completely depend on the hardware in question, given boot processes vary widely between devices. that doesn't seem like an unreasonable approach, but it depends Jul 24 21:28:03 will have a pc on the same network, ethernet connection. Jul 24 22:21:35 RP: I noticed that my git package exploded in size after moving oe-core to https://github.com/openembedded/openembedded-core/commit/bbe0d3e26484f3f347262d40a8a9d415ce21fb43, looking a bit further it seems the hard links were broken when splitting the packages, which resulted in several copies of the git utilities (as it relies on hard links) Jul 24 22:21:55 moving to the previous commit restored back the expected behavior Jul 24 22:22:11 didn't get to debug why yet Jul 24 22:24:45 the inode differences when comparing the files after install in image and after packages-split: https://hastebin.com/yijumecebo.sql Jul 24 22:29:27 rsalveti: clearly there is a bug in the patch as that wasn't intentional Jul 24 22:29:36 rsalveti: I wish we had better tests :/ Jul 24 22:31:39 good at least that it was easy to identify the issue with the git package, as it heavily uses hard links Jul 24 23:40:41 maybe I missed it, but is there any docs on yocto on how it does upgrades? like can I have A/B images on a drive on my device where yocto can upgrade/rollback whole disk images Jul 25 00:16:09 New news from stackoverflow: AM3703 based chip does not boot || Yocto Bluetooth a2dp audio choppy/skipping Jul 25 02:34:16 halfbit: there is no common method, deployment strategies are not hardcoded and is nit out of scope for yocto, but it does provide needed tooling for that e.g. initramfs framework which you would use to achieve the flip/flop update mechanism Jul 25 02:34:54 I have deployed yocto based images in this config too, its fairly straight forward if you stick to tooling and know exactly what you want Jul 25 02:46:34 New news from stackoverflow: Yocto Bluetooth a2dp audio choppy/skipping [on hold] **** ENDING LOGGING AT Wed Jul 25 03:00:02 2018